A high-precision evaluation of ultrasonic detection sensitivity for a micro-crack can be restricted by a corroded rough surface when the surface microtopography is of the same order of magnitude as the crack depth.In this study,a back-surface micro-crack is considered as a research target.A roughness-modified ultrasonic testing model for micro-cracks is established based on a multi-Gaussian beam model and the principle of phase-screen approximation.The echo signals of micro-cracks and noises corresponding to different rough front surfaces and rough back surfaces are obtained based on a reference reflector signal acquired from a two-dimensional simulation model.Further compari-son between the analytical and numerical models shows that the responses of micro-cracks under the effects of dif-ferent corroded rough surfaces can be accurately predicted.The numerical and analytical results show that the echo signal amplitude of the micro-crack decreases significantly with an increase in roughness,whereas the noise ampli-tude slightly increases.Moreover,the effect of the rough front surface on the echo signal of the micro-crack is greater than that of the rough back surface.When the root-mean-square(RMS)height of the surface microtopography is less than 15 μm,the two rough surfaces have less influence on the echo signals detected by a focused transducer with a frequency of 5 MHz and diameter of 6 mm.A method for predicting and evaluating the detection accuracy of micro-cracks under different rough surfaces is proposed by combining the theoretical model and a finite element simulation.Then,a series of rough surface samples containing different micro-cracks are fabricated to experimentally validate the evaluation method.